Akademik Veri Yönetim Sistemi
Circuits, Systems, and Signal Processing, 2026 (SCI-Expanded, Scopus)
This paper proposes an effective proposal method that provides efficient inference for the general problem of recovering of sinusoidal signals from diverse noisy data sources, including non-stationary, correlated, and white noise. For this to be achieved, we propose a novel algorithm that combines a global nonlinear optimization routine based on the Bayesian Inference Reversible Jump Markov Chain Monte Carlo (BI-RJMCMC) method with a range of proposal distributions and simulated annealing (SA). This approach enables a global exploration of the parameters and number of parameters, circumventing the issue of local minima and facilitating a convergence to the modes of the full posterior distribution through an efficient process. The estimation of the parameters can be deduced from these probabilities along with the corresponding variances. Additionally, in instances where the variance of the noise is not known, an estimate of the noise variance can be obtained. A comparison of the results with those obtained using alternative methods, including Bretthorst, Dou, and Hodgson’s methods, has been undertaken. The approach outlined in this paper showed greater feasibility since the selection and estimation of parameters are carried out concurrently in the same sampling algorithm, being more attractive in terms of computational time. Computer simulation results have demonstrated that the proposed Bayesian RJMCMC-SA method is a robust methodology for signal processing in both parameter estimation and model selection.